PROJECT SUMMARY The goal of dose-finding (phase I) cancer clinical trials is to find the optimal dose, or range of doses, deemed safe for patients, which can then be further studied in later phase trials. The National Cancer Institute?s Common Terminology Criteria for Adverse Events (CTCAE) is universally used to define toxicities by type/class and ordinal grade on a scale from 0-5 in terms of increasing severity, where 0 represents no toxicity and 5 represents patient death as a result of toxicity. However, conventional trials that utilize toxicity outcomes generally simplify the toxicity-grading system into a binary outcome, defined by a dose-limiting toxicity (DLT) signifying too toxic of an event in a patient (such that the dose should be reduced), and non-DLT signifying that the dose is not too toxic. This simplification ignores a great deal of toxicity information that could be utilized both clinically, and in the statistical methodology of cancer clinical trials. To better characterize patient toxicity profiles, toxicity scores have been proposed as continuous, composite scores that take into multiple types of toxicity and their entire ordinal grading scale. In a statistical and theoretical framework, the investigator of this project has recently developed an objective, generalizable method to derive a toxicity scoring scheme to be applied to cancer clinical trials, called the Toxicity Score Elicitation Method (TSEM). Aim 1 of this study is to develop and refine the Toxicity Score Elicitation Method for clinical feasibility and application. Working with a group of local oncologists, a computer application prototype will be developed that implements the TSEM in an intuitive user interface for clinicians to use. A survey of oncologist with expertise in early-phase cancer clinical trials will be conducted, where each participating oncologist will be provided a set of hypothetical clinical trial scenarios, and will generate a toxicity scoring scheme for each scenario using the TSEM application. Feedback from clinicians will be collected and used to refine the method and application in its continued development, and results from the toxicity scores will be used to improve upon the TSEMs underlying statistical methodology. Aim 2 will utilize toxicity scores in developing a novel statistical dose-finding method for drug-combination trials that distinguishes attributable toxicities for each agent of a combination therapy. Drug-combination therapies have recently become commonplace in cancer trials, and often combine different classes of drugs together which can lead to complex toxicity profiles. Existing methods fail to account for the possibility of particular toxicities being attributed to a specific drug of a combination, and the traditional binary toxicity endpoint ignores a great deal of information that would be useful in doing so. Using toxicity scores, this research will investigate statistical methods of estimating the probability that a particular toxicity is associated with specific drugs in a combination therapy, and implement this framework into the underlying statistical dose-finding methodology to better identify an optimal dose, or range of doses for further clinical trial evaluation.